Fynsa
Half-jokingly, half-seriously, a few weeks ago I used ChatGPT to write a column for this very space. It’s true that I had to tweak a few things—make the text flow better and cut down on the number of consecutive periods and repeated words—so it wouldn’t sound like a robot wrote it (no pun intended). However, the result was more than acceptable, and my involvement was mainly to add “my own touch” rather than to correct, for example, the substance of the piece. In any case, this doesn’t prove anything, since its performance is still quite poor for other types of tasks, and it continues to falter when it comes to verifying sources that may say contradictory things. For now, at least.
Although this type of experiment has continued (text-generating tools), the focus of the debate on artificial intelligence and ChatGPT may lead the public to believe that they are synonymous and that future development is tied to the problems I described in the previous paragraph. Nothing could be further from the truth. Just as blockchain technology is not just about cryptocurrencies, artificial intelligence is not just a text generator. And the race to be the first to launch the next big tool has meant that a number of questions and debates have not been explored in sufficient depth—or even had the chance to take place at all. For example, consider this scenario: faced with a fortuitous event in which an accident is inevitable and third parties could be affected, whom should an autonomous vehicle prioritize saving?
That is why a group of world leaders (and minor economists at the end of the world), participants in the Future of Life Institute, signed an open letter calling for a pause in large-scale artificial intelligence experiments for at least six months. According to the authors, this pause should be public and verifiable, and if it cannot be implemented quickly, it should be imposed by governments. The reasons are clear: to develop a series of protocols that ensure AI systems operate safely beyond any reasonable doubt; to work with governments and authorities to establish robust institutions capable of overseeing AI, monitoring its progress and growth, and creating “watermarks” to distinguish the real from the synthetic in the event of model leaks, etc. Last but not least, we must secure funding to minimize the social and economic impacts of the drastic changes in people’s lives and jobs that the rapid adoption of AI could cause—changes that could even threaten democracy.
On a personal level, I think it’s time to stop and think about the ethical and moral issues that arise when the decision-maker isn’t a person. What’s more, it may have been a person—the one who designed and programmed the AI—but that person doesn’t necessarily represent the sentiments of society, much less was elected (again, in a democracy) to do so. Let’s return to the question of self-driving cars. Some manufacturers already have an answer for that: the vehicle will always try to minimize harm to its occupants. It may seem obvious, but let me push this idea a little further: what if the occupants are drug traffickers fleeing a crime and, after driving over a slippery patch, head straight toward a group of pregnant women returning from a prenatal yoga class? Perhaps that’s too extreme—let’s consider an easier scenario: A scientist carrying the cure for cancer is in the car, but the vehicle has lost control and is about to run over (unless a quick turn of the steering wheel sends the vehicle into a ravine) the person you love most in the world. Was that decision easier? Is there a “right” answer to these dilemmas? Should the person who programmed the AI—with their biases and experiences—make these decisions, or should it be a body elected by citizens? Which citizens? Is there a global ethical standard, or should it vary by culture, nation, or lifestyle? Remember that this is just one example, since just as AI isn’t just ChatGPT, it isn’t just about autonomous driving either.
Six months might not be much time, but it’s better than nothing. We won’t be able to solve everything—that’s true—but it will give us a chance to think about how we organize what we have and where we want all of this to go from here. Because, at least for me, I don’t want a tyranny ruled by those who run AI companies today—even if that’s not their intention—and certainly not a Skynet.
